Blind spatial interference suppression of DS-CDMA with long spreading sequences

We develop two new algorithms for adaptive antenna array beamforming using DS-CDMA with long spreading codes. While it is assumed that the spreading code and timing for the desired user are known, the adaptation is blind, in that the symbols sent by the desired user are not known. The receiver implements temporal matched filtering (by despreading with respect to the desired user´s spreading waveform) and spatial linear MMSE correlation (after first estimating the desired user´s array response). Our first algorithm blindly estimates the array response of the desired user by computing the fourth order cumulant after despreading using the desired user´s spreading waveform. This and other more complicated algorithms based on higher order statistics (HOS) are not found to be competitive with state of the art blind algorithms based on second order statistics (SOS). Our second algorithm, whose performance is superior to the state of the art, is based on SOS, and exploits a novel interpretation of long spreading codes as a system with a known training sequence at the chip rate (the desired user´s spreading sequence), with the unknown transmitted symbols playing the role of a time varying gain that is piecewise constant